Phonologically Conditioned Phonetic Changes

Josef Fruehwald, University of Pennsylvania
May 1, 2013

Outline

  1. Brief Background
    1. Brief background regarding the conventional wisdom about phonetic changes that I’ll be trying to problemetize.
    2. Brief background on the Philadelphia Neighborhood Corpus.
  2. Two Case studies highlighting the role of phonology in phonetic change
    1. Pre-voiceless /ay/ raising.
    2. Pre-consonantal /ey/ Raising
  3. Conclusions

Background

Conventional wisdom regarding conditioned sound changes:

Natural Physiological / Perceptual Phenomenon

Accumulation of Errors

Phonological Process

e.g. Hyman (1976); Ohala (1981, 1990, inter alia); Blevins (2004); Bermudez-Otero (2007), cf. Baker, Archangelie & Mielke (2011)

Background

Hypothetical: The effect of coronals on adjacent /uw/. (Ohala, 1981)

jawn 

Background

Effect of coronals on /uw/ in Philadelphia.

plot of chunk plot_uws  

Today

I will be presenting an analysis based on the Philadelphia Neighborhood Corpus which problemetizes this conventional wisdom.

Pre-voiceless /ay/ raising.

The opaque interaction between /ay/ raising and /t, d/ flapping is in place from the very beginning.

Pre-consonantal /ey/ raising.

The most phonetically favoring context does not undergo the change.

The Philadelphia Neighborhood Corpus

Philadelphia Row Homes, 1973  
Philadelphia Row Homes, 1973
North Philadelphia Row Homes, 1973
North Philadelphia Row Homes, 1973
North Philadelphia Row Homes, 1973
North Philadelphia Row Homes, 1973
South Philadelphia Row Homes, 2012
South Philadelphia Row Homes, 2012
South Philadelphia Row Homes, 2012  
South Philadelphia Row Homes, 2012

The Philadelphia Neighborhood Corpus

FAVE online suite

Pre-voiceless /ay/ raising

In my own speech […] this short sound, as in night or out, seems to be identical with the vowel of but; contrast with [a] in ride, loud. No such distinction is made in the Philadelphia dialect. – Tucker (1944)

Pre-voiceless /ay/ raising

plot of chunk ays_plot

Pre-voiceless /ay/ raising

Within words, it is only following voiceless segments which condition raising.

plot of chunk plot_ay_internal

Pre-voiceless /ay/ raising

There is no condtioning of /ay/ raising across word boundaries.

plot of chunk plot_ay_final

/ay/ Raising and Opacity

In contemporary Philadelphian English, /ay/ raising applies opaquely with respect to /t/ and /d/ flapping.

write [ɻʌit] [ɻaɪd] ride
writer [ɻʌiɾɚ]
[ɻaɪɾɚ] rider

If /ay/ raising were phonetically conditioned, when did this opacity enter the dialect?

/ay/ Raising and Opacity

with(ayTD_comp, table(FolSeg, Flapping))
##       Flapping
## FolSeg Flap Surface
##      D  328     647
##      T  240    2155

/ay/ Raising and Opacity

Pre-voiceless shortening is commonly appealed to as the phonetic precursor for /ay/ raising (Joos, 1942; Chambers, 1973) pace Thomas & Moreton (2004).

plot of chunk plot_ay_duration

/ay/ Raising and Opacity

plot of chunk ay_flap_plot

/ay/ Raising and Opacity

To see if /ay/ raising always really patterned according to the underlying voicing, we need to do something more complex than linear regression.

plot of chunk ay_rean_plot

/ay/ Raising Modeling

This is the basic model I fit.

 

/ay/ Raising Modeling

surf_d; t_effect; t_flapping; d_flapping ~ b.spline(DOB)

(surf_d; t_effect; t_flapping; d_flapping | Speaker)

(1|Word)

Dur_msec + (Dur_msec | Speaker)

/ay/ Raising Modeling

These are the estimated F1 trajectories from the model with 95% highest posterior density estimate.

plot of chunk plot_ests

/ay/ Modeling

There appears to be a main effect of following flaps being slighly lower, counter to the duration predictions.

plot of chunk ay_flap_effects

/ay/ Modeling

The difference between /t/ and /d/ appears to be identical between the non-flapping and flapping contexts.

plot of chunk t_effects

/ay/ Modeling

The difference between /t/ and /d/ appears to be identical between the non-flapping and flapping contexts.

plot of chunk flap_comp

/ay/ Conclusions

What about analogy?

/ey/ Raising

In long a, as in day, the first element ranges from [æ] to [a]. The resulting diphthong […] [is] treated in nearly the same way in some English dialects, notably Cockney, where indeed the same tendency has been carried still further ([dɑɪ]). (Tucker 1944)

/ey/ Raising

The basic distinction has been previously described as between word-final vs word-internal /ey/.

plot of chunk plot_basic_eys

/ey/ Raising

Preliminary exploratory analysis has found that syllable structure doesn’t matter, just whether the following segment is a consonant (excluding /l/).

plot of chunk plot_eys_internal

/ey/ Raising

But, there is no opacity.

plot of chunk plot_day

/ey/ Raising

And, it is even conditioned across word boundaries.

plot of chunk plot_eys_final

/ey/ Raising

A regression for word internal /ey/ shows that /ey/ followed by /l/ was reliably more advanced than before other consonants, but did not undergo the change.

Decade = (DOB-1900)/10
Diag ~ Decade * FolSeg + (FolSeg | Speaker) + (Decade | Word)
C Estimates Interactions
Intercept 0.63 t = 15.14 -0.55 t = -4.29 FolSeg = V
0.31 t = 2.66 FolSeg = /l/
Decade 0.12 t = 14.9 -0.12 t = -5.22 FolSeg = V × Decade
-0.1 t = -4.18 FolSeg = /l/ × Decade

/ey/ Raising

plot of chunk ey_fit_plot

/ey/ Conclusions

Conclusions

/ay/ raising and /ey/ raising
/ay/ raising
  • Has never applied across word boundaries.
  • Interacts opaquely with flapping.
/ey/ raising
  • Appears to apply across word boundaries.
  • Interacts transparently with affixing of all varieties.

Conclusions

Conclusions

References

References